Computer aided diagnostic methods are more and more used with the extensive integration of electronic and electrical equipment, for instance into motorcars to control engines, braking systems and other on-board systems. For example, the multiplexing technology, used by car manufacturers and suppliers, replaces the numerous hard-wired electrical connections between components by buses. These buses allows for transmission of individual or broadcast messages to and from all components.
The first intent of a diagnostic method is to determine the failure and the action to be accomplished to repair it. The more complex the system for which a diagnostic is required, the more the diagnostic method must be efficient. The use of an efficient method results in saving time, saving money on the parts to be replaced and on skilled people to perform diagnostics. The technician in a garage usually relies on static decision trees for diagnostics. The technician also needs to consult vehicle schematics and parameter readers that he connects to the system to be diagnosed. With this method, except if the technician is particularly expert in a given system, there is no insurance that the appropriate part of the system will be replaced. For complex failures, the technician may use a “trial and error” strategy as he cannot go deeper in the failure analysis.
There is a need for an accurate and efficient computer aided method to save time and to save money on the parts to be replaced while avoiding use of highly skilled technicians to perform diagnostics.
Computer aided methods for failure identification of systems, such as DIAG 2000 from ACTIA, Clip from SAGEM and DIS from Siemens applied to vehicle diagnostic, are available today. They all are implemented with a data repository containing reference information and an engine able to read parameters from the system for which a diagnostic is required. The methods analyze and compute them with data read in the reference data repository to perform the failure identification. According to the type of data stored in the data repository and the algorithm of the engine, the method will have different capacity as for the accuracy and efficiency of the diagnostic, time for processing or the easiness of its adaptation to different systems or its maintenance over time. Basic diagnostic systems use static diagnostic tree similar to yes-no tree while advanced ones use either case based reasoning, model based reasoning or rule based reasoning algorithms which are the known approaches for building Expert systems.
Case based method reasoning consists of matching new problems to “cases” from a historical database and then adapting successful solutions from the past to current situations. Case-based reasoning is a problem solving paradigm that in many respects is fundamentally different from other major AI approaches. Instead of relying solely on general knowledge of a problem domain, or making associations along generalized relationships between problem descriptors and conclusions, CBR is able to utilize the specific knowledge of previously experienced and concrete problem situations (cases). A new problem is solved by finding a similar past case and reusing it in the new problem situation. A second important difference is that CBR also is an approach to incremental, sustained learning, since a new experience is retained each time a problem has been solved, making it immediately available for future problems. When using CBR for a diagnostic determination method, the first step is very fast, it consists in taking a snapshot of a system to be analyzed, which means collection of an amount of parameters characterizing the system at a given time. However, the drawback is that the result is approximate and could be very far from the exact result if the failure problem has not been modeled. This can lead to an error of diagnostic which is not reported as is. The second problem with this method is that the model changes with time and the model accuracy after a while is very difficult to evaluate. This enforces inaccuracy of a diagnostic solution.
In model-based reasoning (MBR), the knowledge base is represented as a set of models of the world rather than a logical formula describing it. Models are parameterized and model-based representation efficiently supports reasoning when varying context information. In order to be operational, with MBR, every object needs to be modeled which may be very costly even if, in the case of an application to a diagnostic method, it is known that some parts or functions cannot be the reason for failure. As the model context needs to be reproduced, with an MBR diagnostic method, the technician in a garage will have to perform a lot of operations to test a specific model. A major problem with MBR applied to a diagnostic method is that the models should evolve with the time to adapt to an evolving complex system, such as a car engine. In this case, the method does not give a way to evaluate the model changes.
Rule-based expert systems are also used. Using a set of assertions, which collectively form the ‘working memory’, and a set of rules that specify how to act on the assertion set, a rule-based system can be created. Rule-based systems are fairly simplistic, consisting of little more than a set of if-then statements, but provide the basis for so-called “expert systems” which are widely used in many fields. The concept of an expert system is this; the knowledge of an expert is encoded into the rule set. When exposed to the same data, the expert system AI will perform in a similar manner to the expert. The main problem with an RBR diagnostic determination method is that the number of rules is so important that the database becomes rapidly unmanageable. As an example, if the system comprises 47,000 parts, the number of rules can grow up to 47,000 multiplied by the number of possible configurations.
Consequently, there is a need for an efficient guided diagnostic method, accurate, realistic in terms of database size and where content can be entered and maintained easily over time.